290 WALTER S. HUNTER 



co-ordinations, except when these are caused to disintegrate 

 rapidly by severe punishment, which was not the case in the 

 present instance, will keep the learning curve high. Pigeon 

 no. 2, e. g., never permitted the average error record to drop 

 below .2. He practically always entered the first cul-de-sac. 

 Several times he passed its entrance and w^ent at least a third 

 of the way through the maze only to return, make the error, and 

 then continue correctly. The persistence of old habits, though, 

 is not the cause of that characteristic of the present curves 

 which we are now discussing, for although pigeons nos. 5 and 8 

 had no old habits carried over from A to lead them astray, their 

 curves are similar to the others. Since, then, the characteristic 



and 

 Errors 



:i 



A 



t 



'" ''^ / 7 IS I 14 ■>■! Trio/, 



Figure 8— Time ( ) and error ( ) graphs for the rotation 



■ tests. A, the last two trials of normal learning; B, rotation of 90°; C, rotation 

 of 270°. 



in question is not due to previous training and since it is not 

 a necessary attribute of the pigeon's learning records — witness 

 Rouse's results and those presented above for labyrinth A — it 

 must be due to a peculiarity of the maze itself, viz., its complexity. 

 This explanation is strengthened very much by the fact that 

 the learning curves for labyrinth C, fig. 9, are of the same nature 

 as those for B. 



What is here presented, then, by way of an hypothesis, is 

 a criterion by which to judge of the complexity of a maze or 

 other problem in relation to an animal's abiHty to learn it. The 

 greater this complexity, the longer will the curve be main- 

 tained at or above a certain point which is determined by a 

 set of fairly easily ascertainable facts and which is usually the 



